Data Center Advisors

So today I got to wondering why on earth would EMC buck their normal trend, having only one enterprise array in the market at a time, with V-Max. This is one of those things that make you go “hmm”. Is there some uncertainty with the shiny new V-Max and keeping DMX around hedges risk? Is there an inherent technical flaw? Hmm. Well this got me to thinking so I reached out to a source who I will call Mr. Scott MacMillan on what his take is as to why the trend has changed. To better understand this we have to have the context of history.

1. From Symm-2 through Symm-4 the core processing technology used was Motorola 68k based and the caching algorithms were hand tuned 68k assembly.
2. As the 68k family was dying off, note that EMC was the last consumer of this processor, EMC picked well the 68k’s successor the PowerPC family. However, oops, those 68k assembly based caching algorithms represented a bit of a problem as they were rather like spooky voodoo black magic and well many of the magicians were well no longer at EMC. So EMC worked with someone who had been through a similar transition — think of a company with a fruit in their logo — and worked on cross compiling from 68k to PowerPC and voila Symm-5.
3. Now was the time to do the transition to DMX or Symm-6. Here was the advent of the mesh architecture — kind of seems like chicken wire to me, but heck — that sported packet-ization of the bus protocol. This was done, drum roll please, yes because well the spooky voodoo black magic could not be disturbed and the hardware was basically tricking the software into thinking that the hardware was just last year’s Symm-5.
4. Next comes that ever death defying stunt, yes that’s right changing the processor technology again to Intel and this time for V-Max. However what to do, the spooky voodoo was still around and Intel assembly is dramatically different from 68k or even PPC? Well presto yes ladies and gentlemen cross compilation and a virtualization container come to the rescue. Further if you notice again there is a customized Link ASIC in front of that Rapid-IO (I like Ken’s Radio-IO, but that is indeed a misspelling) to again trick out the spooky voodoo into thinking that this was a older Symm-5 based system.

So in one long run on sentence, what is that V-Max? The brand-new V-Max is really “68k assembly code cross-compiled to PPC code cross-compiled to x86 code running in a virtual container with custom hardware to support the legacy 68k code”. Mystery solved, EMC has used customized hardware, incantation, and cross compilation to augment their reality so that V-Max could be born. Now I can understand why they would keep both product lines around: lack of confidence. Hey this seems like a feat worthy of a Salvador Dali painting.

UPDATE

Note that Barry correctly pointed out that this is M.C. Escher’s “Relativity“.  So I’m corrected in this point.  I also have to say when I look at the very beginning of the Wikipedia entry I see the following quote: “It depicts a world in which the normal laws of gravity do not apply”.  In this I agree with Barry, this is a great representation of EMC as they tend to make up their own Augmented Reality where they interpret and generate their own “laws of gravity”.  Sorry Barry I could not resist, and the proper term of response is touché.

Tweet

Posted in Tech Talk 9 Comments »

Comments (9 )

Post Comment

Very creative, but hardly the reality. I guess V-Max is having the desired effect on USP-V sales if you have to stoop to this level of misreputation.

Product transitions are never “instant”, thus DMX4 remains in the active product line for many of the same reasons that Hitachi continued to bid/sell the old TagmaSnore (USP) for more than 12 months after the introduction of the USP-V.

Props for your imagination though.

By the way – that’s not a Dali – it is an MC Escher etching titled “Relativity” (notice the “MCE” in the top left corner).

It is thus an appropriate image to use in reference to EMC^2

Barry,

Then what is the reality both on the very specific architectural comments Michael made? What about the actual sales? (readers, please see my blog entry http://blogs.hds.com/christophe/2009/07/is-vmax-stuck-on-the-runway.html ) .

Also, after asking around at HDS, I don’t believe we have lost one deal to VMax. One of our sales guys said you should bundle it with InVista (That was not very nice).

It’s definitely not an instant transition as you said.

Christophe -

EMC PR informs me that EMC does not make public statements about unit sales of any of our products, so I am not able to provide specifics for V-Max sales. But I can say this:

V-Max sales are meeting or exceeding the revenue and unit plans.

And whether or not you “believe” that the USP-V is losing sales to V-Max doesn’t really matter, now does it? Are you sure that you know about all HDS, HP and Sun sales engagements?

As to the architectural questions that you say sre “very specific” – I’ll answer the only 3 specific questions I see:

Is EMC keeping the DMX4 to hedge risk? No
Is there an inherent technical flaw? No
What is that V-Max? The latest generation of the #1 market share high-end storage platform, that’s what.

Michael –

There is very little assembly code in the latest generations of Enginuity for both DMX and V-Max, and what there is has been written-for-purpose, not cross-compiled. The majority of the code is written in high-level languages, often hand-optimized for the target processor complex and global memory interconnect.

The V-Max ASIC indeed enables the Intel quad-core Xeon processors to transparently access both local and remote memory as one contiguous memory address space, simplifying (and shortening) code paths, allowing every processor to access well in excess of the 1TB max global memory of today’s largest V-Max.

And the “voodoo” you try to invoke as a negative is actually the architectural foundation of Symmetrix. Yes, indeed, every major generation is evolution of core architecture, optimized to take advantage in advances in technology and for different hardware platforms.

You can make fun of it if you like, but that architecture has helped keep 8 generations of Symmetrix #1 in high-end market share for the past 20 years.

Barry which code is written in assembly?

Micheal – what, do you want module names?

I don’t think I can answer that one …

Barry, I’ll make the change for the post shortly, thanks for catching the incorrectly referenced work of art. Note in a joking jousty point when I looked up the work for a giggle I found that one of the things which is used to describe the work is that of a “world where the laws of gravity do not apply.” You are right I think that this nicely describes EMC’s approach to the world!

Barry please get your facts straight. I’ll go against my compatriots a little bit, EMC and Hitachi routinely battle for #1 share in the Enterprise Storage segment. Sometimes Hitachi is #1 and EMC #2 and vice versa, but what we can definitely agree upon is that IBM is well below both of us!!!

Post a Comment